地球物理学进展 ›› 2018, Vol. 33 ›› Issue (4): 1622-1628.doi: 10.6038/pg2018BB0346

• 应用地球物理学Ⅰ(油气及金属矿产地球物理勘探) • 上一篇    下一篇

基于自相关函数特性的CEEMD全局阈值去噪方法研究

杨涛,乐友喜,曾贤德,蔡俊雄,曾勉   

  1. 中国石油大学(华东)地球科学与技术学院,山东青岛 266580
  • 收稿日期:2017-10-01 修回日期:2018-06-10 出版日期:2018-08-20 发布日期:2018-09-14
  • 作者简介:杨涛,男,1993年生,河南驻马店人,研究生,主要从事地震资料解释工作.(E-mail: 765999768@qq.com)

Research on CEEMD global threshold denoising method based on autocorrelation function

YANG Tao,YUE You-xi,ZENG Xian-de,CAI Jun-xiong,ZENG Mian   

  1. China University of Petroleum (East China) School of Geosciences, Shandong Qingdao 266580, China
  • Received:2017-10-01 Revised:2018-06-10 Online:2018-08-20 Published:2018-09-14

摘要:

完备总体经验模态分解 (CEEMD)克服了经验模态分解(EMD)的模态混叠问题,依据信号自身的特点,将待分析的复杂信号分解为一系列不同尺度的固有模态函数(IMF)的子信号,且各IMF分量的频率由高到低依次排列,是一种适用于分析处理非线性非平稳信号的强大的信号分析技术.地震资料中的随机噪声一般属于高频率的信号,在CEEMD中往往分布在前几个高频IMF分量,本文针对基于CEEMD的分频去噪和基于CEEMD的小波阈值去噪等方法的不足,在前人基于EMD阈值去噪的基础上设计了自相关函数统计特性与CEEMD全局阈值联合去噪方法.该方法先对CEEMD分解的若干个模态分量进行自相关,寻找到噪声主导模态和信号主导模态,然后利用设计的全局阈值对噪声主导模态进行去噪,最后将处理后和未处理的固有模态函数进行重构,得到最终的去噪结果.模型试算和实际地震资料处理都验证了此方法在提高信噪比,保留原信号高频有效成分和弱信号信息上的有效性.

关键词: 自相关特性, 分频去噪, 全局阈值, 小波阈值去噪, 完备总体经验模态分解 (CEEMD)

Abstract:

Complete Empirical Empirical Mode Decomposition(CEEMD) overcomes the modal aliasing problem of Empirical Mode Decomposition(EMD). According to the characteristics of the signal itself, the complex signal to be analyzed is decomposed into a series of sub-signals of the Inherent Modal Functions(IMF) of different scales, and the frequency of each component is arranged from high to low order, which is a kind of nonlinear non-stationary signal of the powerful signal analysis technology. The random noise is generally a high frequency signal in seismic data and is often distributed in the first few high frequency IMF components in the CEEMD decomposition. This article aims at the lack of frequency denoising based on CEEMD and wavelet threshold denoising based on CEEMD, statistical properties of autocorrelation functions and global threshold based on CEEMD jointed denoising method is designed on the basis of the previous threshold denoising based on EMD.Firstly,a number of modal components of CEEMD decomposition are autocorreved to find the dominant mode of noise and the dominant mode of signal. Secondly the global threshold is used to denoise the dominant mode of noise. Finally, the processed and unprocessed intrinsic modal functions are reconstructed to obtain the final denoising result. Both the model trial and the actual seismic data processing verify the effectiveness of this method in improving the signal-to-noise ratio, preserving the high frequency active components of original signal and the information of weak signal.

Key words: autocorrelation characteristics, frequency division denoising, global threshold, wavelet threshold denoising, Complete Empirical Empirical Mode Decomposition(CEEMD)

中图分类号: